xref: /openbsd-src/sys/netinet/udp_usrreq.c (revision db3296cf5c1dd9058ceecc3a29fe4aaa0bd26000)
1 /*	$OpenBSD: udp_usrreq.c,v 1.91 2003/07/09 22:03:16 itojun Exp $	*/
2 /*	$NetBSD: udp_usrreq.c,v 1.28 1996/03/16 23:54:03 christos Exp $	*/
3 
4 /*
5  * Copyright (c) 1982, 1986, 1988, 1990, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  *
32  *	@(#)COPYRIGHT	1.1 (NRL) 17 January 1995
33  *
34  * NRL grants permission for redistribution and use in source and binary
35  * forms, with or without modification, of the software and documentation
36  * created at NRL provided that the following conditions are met:
37  *
38  * 1. Redistributions of source code must retain the above copyright
39  *    notice, this list of conditions and the following disclaimer.
40  * 2. Redistributions in binary form must reproduce the above copyright
41  *    notice, this list of conditions and the following disclaimer in the
42  *    documentation and/or other materials provided with the distribution.
43  * 3. All advertising materials mentioning features or use of this software
44  *    must display the following acknowledgements:
45  * 	This product includes software developed by the University of
46  * 	California, Berkeley and its contributors.
47  * 	This product includes software developed at the Information
48  * 	Technology Division, US Naval Research Laboratory.
49  * 4. Neither the name of the NRL nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
54  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
55  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
56  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
57  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
58  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
59  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
60  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
61  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
62  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
63  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
64  *
65  * The views and conclusions contained in the software and documentation
66  * are those of the authors and should not be interpreted as representing
67  * official policies, either expressed or implied, of the US Naval
68  * Research Laboratory (NRL).
69  */
70 
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mbuf.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/sysctl.h>
78 
79 #include <net/if.h>
80 #include <net/route.h>
81 
82 #include <netinet/in.h>
83 #include <netinet/in_systm.h>
84 #include <netinet/in_var.h>
85 #include <netinet/ip.h>
86 #include <netinet/in_pcb.h>
87 #include <netinet/ip_var.h>
88 #include <netinet/ip_icmp.h>
89 #include <netinet/udp.h>
90 #include <netinet/udp_var.h>
91 
92 #ifdef INET6
93 #ifndef INET
94 #include <netinet/in.h>
95 #endif
96 #include <netinet6/ip6protosw.h>
97 
98 extern int ip6_defhlim;
99 #endif /* INET6 */
100 
101 /*
102  * UDP protocol implementation.
103  * Per RFC 768, August, 1980.
104  */
105 int	udpcksum = 1;
106 
107 struct	inpcbtable udbtable;
108 struct	udpstat udpstat;
109 
110 static	void udp_detach(struct inpcb *);
111 static	void udp_notify(struct inpcb *, int);
112 static	struct mbuf *udp_saveopt(caddr_t, int, int);
113 
114 #ifndef UDBHASHSIZE
115 #define	UDBHASHSIZE	128
116 #endif
117 int	udbhashsize = UDBHASHSIZE;
118 
119 /* from in_pcb.c */
120 extern	struct baddynamicports baddynamicports;
121 
122 void
123 udp_init()
124 {
125 	in_pcbinit(&udbtable, udbhashsize);
126 }
127 
128 #ifdef INET6
129 int
130 udp6_input(mp, offp, proto)
131 	struct mbuf **mp;
132 	int *offp, proto;
133 {
134 	struct mbuf *m = *mp;
135 
136 #if defined(NFAITH) && 0 < NFAITH
137 	if (m->m_pkthdr.rcvif) {
138 		if (m->m_pkthdr.rcvif->if_type == IFT_FAITH) {
139 			/* XXX send icmp6 host/port unreach? */
140 			m_freem(m);
141 			return IPPROTO_DONE;
142 		}
143 	}
144 #endif
145 
146 	udp_input(m, *offp, proto);
147 	return IPPROTO_DONE;
148 }
149 #endif
150 
151 void
152 udp_input(struct mbuf *m, ...)
153 {
154 	register struct ip *ip;
155 	register struct udphdr *uh;
156 	register struct inpcb *inp;
157 	struct mbuf *opts = 0;
158 	struct ip save_ip;
159 	int iphlen, len;
160 	va_list ap;
161 	u_int16_t savesum;
162 	union {
163 		struct sockaddr sa;
164 		struct sockaddr_in sin;
165 #ifdef INET6
166 		struct sockaddr_in6 sin6;
167 #endif /* INET6 */
168 	} srcsa, dstsa;
169 #ifdef INET6
170 	struct ip6_hdr *ip6;
171 #endif /* INET6 */
172 #ifdef IPSEC
173 	struct m_tag *mtag;
174 	struct tdb_ident *tdbi;
175 	struct tdb *tdb;
176 	int error, s;
177 #endif /* IPSEC */
178 
179 	va_start(ap, m);
180 	iphlen = va_arg(ap, int);
181 	va_end(ap);
182 
183 	udpstat.udps_ipackets++;
184 
185 	switch (mtod(m, struct ip *)->ip_v) {
186 	case 4:
187 		ip = mtod(m, struct ip *);
188 #ifdef INET6
189 		ip6 = NULL;
190 #endif /* INET6 */
191 		srcsa.sa.sa_family = AF_INET;
192 		break;
193 #ifdef INET6
194 	case 6:
195 		ip = NULL;
196 		ip6 = mtod(m, struct ip6_hdr *);
197 		srcsa.sa.sa_family = AF_INET6;
198 		break;
199 #endif /* INET6 */
200 	default:
201 		goto bad;
202 	}
203 
204 	IP6_EXTHDR_GET(uh, struct udphdr *, m, iphlen, sizeof(struct udphdr));
205 	if (!uh) {
206 		udpstat.udps_hdrops++;
207 		return;
208 	}
209 
210 	/* Check for illegal destination port 0 */
211 	if (uh->uh_dport == 0) {
212 		udpstat.udps_noport++;
213 		goto bad;
214 	}
215 
216 	/*
217 	 * Make mbuf data length reflect UDP length.
218 	 * If not enough data to reflect UDP length, drop.
219 	 */
220 	len = ntohs((u_int16_t)uh->uh_ulen);
221 	if (ip) {
222 		if (m->m_pkthdr.len - iphlen != len) {
223 			if (len > (m->m_pkthdr.len - iphlen) ||
224 			    len < sizeof(struct udphdr)) {
225 				udpstat.udps_badlen++;
226 				goto bad;
227 			}
228 			m_adj(m, len - (m->m_pkthdr.len - iphlen));
229 		}
230 	}
231 #ifdef INET6
232 	else if (ip6) {
233 		/* jumbograms */
234 		if (len == 0 && m->m_pkthdr.len - iphlen > 0xffff)
235 			len = m->m_pkthdr.len - iphlen;
236 		if (len != m->m_pkthdr.len - iphlen) {
237 			udpstat.udps_badlen++;
238 			goto bad;
239 		}
240 	}
241 #endif
242 	else /* shouldn't happen */
243 		goto bad;
244 
245 	/*
246 	 * Save a copy of the IP header in case we want restore it
247 	 * for sending an ICMP error message in response.
248 	 */
249 	if (ip)
250 		save_ip = *ip;
251 
252 	/*
253 	 * Checksum extended UDP header and data.
254 	 * from W.R.Stevens: check incoming udp cksums even if
255 	 *	udpcksum is not set.
256 	 */
257 	savesum = uh->uh_sum;
258 #ifdef INET6
259 	if (ip6) {
260 		/* Be proactive about malicious use of IPv4 mapped address */
261 		if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
262 		    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
263 			/* XXX stat */
264 			goto bad;
265 		}
266 
267 		/*
268 		 * In IPv6, the UDP checksum is ALWAYS used.
269 		 */
270 		if ((uh->uh_sum = in6_cksum(m, IPPROTO_UDP, iphlen, len))) {
271 			udpstat.udps_badsum++;
272 			goto bad;
273 		}
274 	} else
275 #endif /* INET6 */
276 	if (uh->uh_sum) {
277 		if ((m->m_pkthdr.csum & M_UDP_CSUM_IN_OK) == 0) {
278 			if (m->m_pkthdr.csum & M_UDP_CSUM_IN_BAD) {
279 				udpstat.udps_badsum++;
280 				udpstat.udps_inhwcsum++;
281 				m_freem(m);
282 				return;
283 			}
284 
285 			bzero(((struct ipovly *)ip)->ih_x1,
286 			    sizeof ((struct ipovly *)ip)->ih_x1);
287 			((struct ipovly *)ip)->ih_len = uh->uh_ulen;
288 
289 			if ((uh->uh_sum = in_cksum(m, len +
290 			    sizeof (struct ip))) != 0) {
291 				udpstat.udps_badsum++;
292 				m_freem(m);
293 				return;
294 			}
295 		} else {
296 			m->m_pkthdr.csum &= ~M_UDP_CSUM_IN_OK;
297 			udpstat.udps_inhwcsum++;
298 		}
299 	} else
300 		udpstat.udps_nosum++;
301 
302 	switch (srcsa.sa.sa_family) {
303 	case AF_INET:
304 		bzero(&srcsa, sizeof(struct sockaddr_in));
305 		srcsa.sin.sin_len = sizeof(struct sockaddr_in);
306 		srcsa.sin.sin_family = AF_INET;
307 		srcsa.sin.sin_port = uh->uh_sport;
308 		srcsa.sin.sin_addr = ip->ip_src;
309 
310 		bzero(&dstsa, sizeof(struct sockaddr_in));
311 		dstsa.sin.sin_len = sizeof(struct sockaddr_in);
312 		dstsa.sin.sin_family = AF_INET;
313 		dstsa.sin.sin_port = uh->uh_dport;
314 		dstsa.sin.sin_addr = ip->ip_dst;
315 		break;
316 #ifdef INET6
317 	case AF_INET6:
318 		bzero(&srcsa, sizeof(struct sockaddr_in6));
319 		srcsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
320 		srcsa.sin6.sin6_family = AF_INET6;
321 		srcsa.sin6.sin6_port = uh->uh_sport;
322 #if 0 /*XXX inbound flowinfo */
323 		srcsa.sin6.sin6_flowinfo = htonl(0x0fffffff) & ip6->ip6_flow;
324 #endif
325 		/* KAME hack: recover scopeid */
326 		(void)in6_recoverscope(&srcsa.sin6, &ip6->ip6_src,
327 		    m->m_pkthdr.rcvif);
328 
329 		bzero(&dstsa, sizeof(struct sockaddr_in6));
330 		dstsa.sin6.sin6_len = sizeof(struct sockaddr_in6);
331 		dstsa.sin6.sin6_family = AF_INET6;
332 		dstsa.sin6.sin6_port = uh->uh_dport;
333 		/* KAME hack: recover scopeid */
334 		(void)in6_recoverscope(&dstsa.sin6, &ip6->ip6_dst,
335 		    m->m_pkthdr.rcvif);
336 		break;
337 #endif /* INET6 */
338 	}
339 
340 #ifdef INET6
341 	if ((ip6 && IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) ||
342 	    (ip && IN_MULTICAST(ip->ip_dst.s_addr)) ||
343 	    (ip && in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))) {
344 #else /* INET6 */
345 	if (IN_MULTICAST(ip->ip_dst.s_addr) ||
346 	    in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
347 #endif /* INET6 */
348 		struct socket *last;
349 		/*
350 		 * Deliver a multicast or broadcast datagram to *all* sockets
351 		 * for which the local and remote addresses and ports match
352 		 * those of the incoming datagram.  This allows more than
353 		 * one process to receive multi/broadcasts on the same port.
354 		 * (This really ought to be done for unicast datagrams as
355 		 * well, but that would cause problems with existing
356 		 * applications that open both address-specific sockets and
357 		 * a wildcard socket listening to the same port -- they would
358 		 * end up receiving duplicates of every unicast datagram.
359 		 * Those applications open the multiple sockets to overcome an
360 		 * inadequacy of the UDP socket interface, but for backwards
361 		 * compatibility we avoid the problem here rather than
362 		 * fixing the interface.  Maybe 4.5BSD will remedy this?)
363 		 */
364 
365 		iphlen += sizeof(struct udphdr);
366 
367 		/*
368 		 * Locate pcb(s) for datagram.
369 		 * (Algorithm copied from raw_intr().)
370 		 */
371 		last = NULL;
372 		for (inp = udbtable.inpt_queue.cqh_first;
373 		    inp != (struct inpcb *)&udbtable.inpt_queue;
374 		    inp = inp->inp_queue.cqe_next) {
375 #ifdef INET6
376 			/* don't accept it if AF does not match */
377 			if (ip6 && !(inp->inp_flags & INP_IPV6))
378 				continue;
379 			if (!ip6 && (inp->inp_flags & INP_IPV6))
380 				continue;
381 #endif
382 			if (inp->inp_lport != uh->uh_dport)
383 				continue;
384 #ifdef INET6
385 			if (ip6) {
386 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_laddr6))
387 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_laddr6,
388 					    &ip6->ip6_dst))
389 						continue;
390 			} else
391 #endif /* INET6 */
392 			if (inp->inp_laddr.s_addr != INADDR_ANY) {
393 				if (inp->inp_laddr.s_addr !=
394 				    ip->ip_dst.s_addr)
395 					continue;
396 			}
397 #ifdef INET6
398 			if (ip6) {
399 				if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6))
400 					if (!IN6_ARE_ADDR_EQUAL(&inp->inp_faddr6,
401 					    &ip6->ip6_src) ||
402 					    inp->inp_fport != uh->uh_sport)
403 						continue;
404 			} else
405 #endif /* INET6 */
406 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
407 				if (inp->inp_faddr.s_addr !=
408 				    ip->ip_src.s_addr ||
409 				    inp->inp_fport != uh->uh_sport)
410 					continue;
411 			}
412 
413 			if (last != NULL) {
414 				struct mbuf *n;
415 
416 				if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
417 					opts = NULL;
418 #ifdef INET6
419 					if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
420 						ip6_savecontrol(inp, &opts, ip6, n);
421 #endif /* INET6 */
422 					m_adj(n, iphlen);
423 					if (sbappendaddr(&last->so_rcv,
424 					    &srcsa.sa, n, opts) == 0) {
425 						m_freem(n);
426 						if (opts)
427 							m_freem(opts);
428 						udpstat.udps_fullsock++;
429 					} else
430 						sorwakeup(last);
431 					opts = NULL;
432 				}
433 			}
434 			last = inp->inp_socket;
435 			/*
436 			 * Don't look for additional matches if this one does
437 			 * not have either the SO_REUSEPORT or SO_REUSEADDR
438 			 * socket options set.  This heuristic avoids searching
439 			 * through all pcbs in the common case of a non-shared
440 			 * port.  It * assumes that an application will never
441 			 * clear these options after setting them.
442 			 */
443 			if ((last->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
444 				break;
445 		}
446 
447 		if (last == NULL) {
448 			/*
449 			 * No matching pcb found; discard datagram.
450 			 * (No need to send an ICMP Port Unreachable
451 			 * for a broadcast or multicast datgram.)
452 			 */
453 			udpstat.udps_noportbcast++;
454 			goto bad;
455 		}
456 
457 		opts = NULL;
458 #ifdef INET6
459 		if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
460 			ip6_savecontrol(inp, &opts, ip6, m);
461 #endif /* INET6 */
462 		m_adj(m, iphlen);
463 		if (sbappendaddr(&last->so_rcv,
464 		    &srcsa.sa, m, opts) == 0) {
465 			udpstat.udps_fullsock++;
466 			goto bad;
467 		}
468 		sorwakeup(last);
469 		return;
470 	}
471 	/*
472 	 * Locate pcb for datagram.
473 	 */
474 #ifdef INET6
475 	if (ip6)
476 		inp = in6_pcbhashlookup(&udbtable, &ip6->ip6_src, uh->uh_sport,
477 		    &ip6->ip6_dst, uh->uh_dport);
478 	else
479 #endif /* INET6 */
480 	inp = in_pcbhashlookup(&udbtable, ip->ip_src, uh->uh_sport,
481 	    ip->ip_dst, uh->uh_dport);
482 	if (inp == 0) {
483 		++udpstat.udps_pcbhashmiss;
484 #ifdef INET6
485 		if (ip6) {
486 			inp = in_pcblookup(&udbtable,
487 			    (struct in_addr *)&(ip6->ip6_src),
488 			    uh->uh_sport, (struct in_addr *)&(ip6->ip6_dst),
489 			    uh->uh_dport, INPLOOKUP_WILDCARD | INPLOOKUP_IPV6);
490 		} else
491 #endif /* INET6 */
492 		inp = in_pcblookup(&udbtable, &ip->ip_src, uh->uh_sport,
493 		    &ip->ip_dst, uh->uh_dport, INPLOOKUP_WILDCARD);
494 		if (inp == 0) {
495 			udpstat.udps_noport++;
496 			if (m->m_flags & (M_BCAST | M_MCAST)) {
497 				udpstat.udps_noportbcast++;
498 				goto bad;
499 			}
500 #ifdef INET6
501 			if (ip6) {
502 				icmp6_error(m, ICMP6_DST_UNREACH,
503 				    ICMP6_DST_UNREACH_NOPORT,0);
504 			} else
505 #endif /* INET6 */
506 			{
507 				*ip = save_ip;
508 				uh->uh_sum = savesum;
509 				icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT,
510 				    0, 0);
511 			}
512 			return;
513 		}
514 	}
515 
516 #ifdef IPSEC
517 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_IN_DONE, NULL);
518 	s = splnet();
519 	if (mtag != NULL) {
520 		tdbi = (struct tdb_ident *)(mtag + 1);
521 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
522 	} else
523 		tdb = NULL;
524 	ipsp_spd_lookup(m, srcsa.sa.sa_family, iphlen, &error,
525 	    IPSP_DIRECTION_IN, tdb, inp);
526 	if (error) {
527 		splx(s);
528 		goto bad;
529 	}
530 
531 	/* Latch SA only if the socket is connected */
532 	if (inp->inp_tdb_in != tdb &&
533 	    (inp->inp_socket->so_state & SS_ISCONNECTED)) {
534 		if (tdb) {
535 			tdb_add_inp(tdb, inp, 1);
536 			if (inp->inp_ipo == NULL) {
537 				inp->inp_ipo = ipsec_add_policy(inp,
538 				    srcsa.sa.sa_family, IPSP_DIRECTION_OUT);
539 				if (inp->inp_ipo == NULL) {
540 					splx(s);
541 					goto bad;
542 				}
543 			}
544 			if (inp->inp_ipo->ipo_dstid == NULL &&
545 			    tdb->tdb_srcid != NULL) {
546 				inp->inp_ipo->ipo_dstid = tdb->tdb_srcid;
547 				tdb->tdb_srcid->ref_count++;
548 			}
549 			if (inp->inp_ipsec_remotecred == NULL &&
550 			    tdb->tdb_remote_cred != NULL) {
551 				inp->inp_ipsec_remotecred =
552 				    tdb->tdb_remote_cred;
553 				tdb->tdb_remote_cred->ref_count++;
554 			}
555 			if (inp->inp_ipsec_remoteauth == NULL &&
556 			    tdb->tdb_remote_auth != NULL) {
557 				inp->inp_ipsec_remoteauth =
558 				    tdb->tdb_remote_auth;
559 				tdb->tdb_remote_auth->ref_count++;
560 			}
561 		} else { /* Just reset */
562 			TAILQ_REMOVE(&inp->inp_tdb_in->tdb_inp_in, inp,
563 			    inp_tdb_in_next);
564 			inp->inp_tdb_in = NULL;
565 		}
566 	}
567 	splx(s);
568 #endif /*IPSEC */
569 
570 	opts = NULL;
571 #ifdef INET6
572 	if (ip6 && (inp->inp_flags & IN6P_CONTROLOPTS))
573 		ip6_savecontrol(inp, &opts, ip6, m);
574 #endif /* INET6 */
575 	if (ip && (inp->inp_flags & INP_CONTROLOPTS)) {
576 		struct mbuf **mp = &opts;
577 
578 		if (inp->inp_flags & INP_RECVDSTADDR) {
579 			*mp = udp_saveopt((caddr_t) &ip->ip_dst,
580 			    sizeof(struct in_addr), IP_RECVDSTADDR);
581 			if (*mp)
582 				mp = &(*mp)->m_next;
583 		}
584 #ifdef notyet
585 		/* options were tossed above */
586 		if (inp->inp_flags & INP_RECVOPTS) {
587 			*mp = udp_saveopt((caddr_t) opts_deleted_above,
588 			    sizeof(struct in_addr), IP_RECVOPTS);
589 			if (*mp)
590 				mp = &(*mp)->m_next;
591 		}
592 		/* ip_srcroute doesn't do what we want here, need to fix */
593 		if (inp->inp_flags & INP_RECVRETOPTS) {
594 			*mp = udp_saveopt((caddr_t) ip_srcroute(),
595 			    sizeof(struct in_addr), IP_RECVRETOPTS);
596 			if (*mp)
597 				mp = &(*mp)->m_next;
598 		}
599 #endif
600 	}
601 	iphlen += sizeof(struct udphdr);
602 	m_adj(m, iphlen);
603 	if (sbappendaddr(&inp->inp_socket->so_rcv, &srcsa.sa, m, opts) == 0) {
604 		udpstat.udps_fullsock++;
605 		goto bad;
606 	}
607 	sorwakeup(inp->inp_socket);
608 	return;
609 bad:
610 	m_freem(m);
611 	if (opts)
612 		m_freem(opts);
613 }
614 
615 /*
616  * Create a "control" mbuf containing the specified data
617  * with the specified type for presentation with a datagram.
618  */
619 struct mbuf *
620 udp_saveopt(p, size, type)
621 	caddr_t p;
622 	register int size;
623 	int type;
624 {
625 	register struct cmsghdr *cp;
626 	struct mbuf *m;
627 
628 	if ((m = m_get(M_DONTWAIT, MT_CONTROL)) == NULL)
629 		return ((struct mbuf *) NULL);
630 	cp = (struct cmsghdr *) mtod(m, struct cmsghdr *);
631 	bcopy(p, CMSG_DATA(cp), size);
632 	size = CMSG_LEN(size);
633 	m->m_len = size;
634 	cp->cmsg_len = size;
635 	cp->cmsg_level = IPPROTO_IP;
636 	cp->cmsg_type = type;
637 	return (m);
638 }
639 
640 /*
641  * Notify a udp user of an asynchronous error;
642  * just wake up so that he can collect error status.
643  */
644 static void
645 udp_notify(inp, errno)
646 	register struct inpcb *inp;
647 	int errno;
648 {
649 	inp->inp_socket->so_error = errno;
650 	sorwakeup(inp->inp_socket);
651 	sowwakeup(inp->inp_socket);
652 }
653 
654 #ifdef INET6
655 void
656 udp6_ctlinput(cmd, sa, d)
657 	int cmd;
658 	struct sockaddr *sa;
659 	void *d;
660 {
661 	struct udphdr uh;
662 	struct sockaddr_in6 sa6;
663 	register struct ip6_hdr *ip6;
664 	struct mbuf *m;
665 	int off;
666 	void *cmdarg;
667 	struct ip6ctlparam *ip6cp = NULL;
668 	struct in6_addr finaldst;
669 	struct udp_portonly {
670 		u_int16_t uh_sport;
671 		u_int16_t uh_dport;
672 	} *uhp;
673 	void (*notify)(struct inpcb *, int) = udp_notify;
674 
675 	if (sa == NULL)
676 		return;
677 	if (sa->sa_family != AF_INET6 ||
678 	    sa->sa_len != sizeof(struct sockaddr_in6))
679 		return;
680 
681 	if ((unsigned)cmd >= PRC_NCMDS)
682 		return;
683 	if (PRC_IS_REDIRECT(cmd))
684 		notify = in_rtchange, d = NULL;
685 	else if (cmd == PRC_HOSTDEAD)
686 		d = NULL;
687 	else if (cmd == PRC_MSGSIZE)
688 		; /* special code is present, see below */
689 	else if (inet6ctlerrmap[cmd] == 0)
690 		return;
691 
692 	/* if the parameter is from icmp6, decode it. */
693 	if (d != NULL) {
694 		ip6cp = (struct ip6ctlparam *)d;
695 		m = ip6cp->ip6c_m;
696 		ip6 = ip6cp->ip6c_ip6;
697 		off = ip6cp->ip6c_off;
698 		cmdarg = ip6cp->ip6c_cmdarg;
699 	} else {
700 		m = NULL;
701 		ip6 = NULL;
702 		cmdarg = NULL;
703 		/* XXX: translate addresses into internal form */
704 		sa6 = *(struct sockaddr_in6 *)sa;
705 #ifndef SCOPEDROUTING
706 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
707 			/* should be impossible */
708 			return;
709 		}
710 #endif
711 	}
712 
713 	if (ip6cp && ip6cp->ip6c_finaldst) {
714 		bzero(&sa6, sizeof(sa6));
715 		sa6.sin6_family = AF_INET6;
716 		sa6.sin6_len = sizeof(sa6);
717 		sa6.sin6_addr = *ip6cp->ip6c_finaldst;
718 		/* XXX: assuming M is valid in this case */
719 		sa6.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
720 		    ip6cp->ip6c_finaldst);
721 #ifndef SCOPEDROUTING
722 		if (in6_embedscope(ip6cp->ip6c_finaldst, &sa6, NULL, NULL)) {
723 			/* should be impossible */
724 			return;
725 		}
726 #endif
727 	} else {
728 		/* XXX: translate addresses into internal form */
729 		sa6 = *(struct sockaddr_in6 *)sa;
730 #ifndef SCOPEDROUTING
731 		if (in6_embedscope(&sa6.sin6_addr, &sa6, NULL, NULL)) {
732 			/* should be impossible */
733 			return;
734 		}
735 #endif
736 	}
737 
738 	if (ip6) {
739 		/*
740 		 * XXX: We assume that when IPV6 is non NULL,
741 		 * M and OFF are valid.
742 		 */
743 		struct sockaddr_in6 sa6_src;
744 
745 		/* check if we can safely examine src and dst ports */
746 		if (m->m_pkthdr.len < off + sizeof(*uhp))
747 			return;
748 
749 		bzero(&uh, sizeof(uh));
750 		m_copydata(m, off, sizeof(*uhp), (caddr_t)&uh);
751 
752 		bzero(&sa6_src, sizeof(sa6_src));
753 		sa6_src.sin6_family = AF_INET6;
754 		sa6_src.sin6_len = sizeof(sa6_src);
755 		sa6_src.sin6_addr = ip6->ip6_src;
756 		sa6_src.sin6_scope_id = in6_addr2scopeid(m->m_pkthdr.rcvif,
757 		    &ip6->ip6_src);
758 #ifndef SCOPEDROUTING
759 		if (in6_embedscope(&sa6_src.sin6_addr, &sa6_src, NULL, NULL)) {
760 			/* should be impossible */
761 			return;
762 		}
763 #endif
764 
765 		if (cmd == PRC_MSGSIZE) {
766 			int valid = 0;
767 
768 			/*
769 			 * Check to see if we have a valid UDP socket
770 			 * corresponding to the address in the ICMPv6 message
771 			 * payload.
772 			 */
773 			if (in6_pcbhashlookup(&udbtable, &finaldst,
774 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport))
775 				valid = 1;
776 			else if (in_pcblookup(&udbtable, &sa6.sin6_addr,
777 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport,
778 			    INPLOOKUP_IPV6))
779 				valid = 1;
780 #if 0
781 			/*
782 			 * As the use of sendto(2) is fairly popular,
783 			 * we may want to allow non-connected pcb too.
784 			 * But it could be too weak against attacks...
785 			 * We should at least check if the local address (= s)
786 			 * is really ours.
787 			 */
788 			else if (in_pcblookup(&udbtable, &sa6.sin6_addr,
789 			    uh.uh_dport, &sa6_src.sin6_addr, uh.uh_sport,
790 			    INPLOOKUP_WILDCARD | INPLOOKUP_IPV6))
791 				valid = 1;
792 #endif
793 
794 			/*
795 			 * Depending on the value of "valid" and routing table
796 			 * size (mtudisc_{hi,lo}wat), we will:
797 			 * - recalcurate the new MTU and create the
798 			 *   corresponding routing entry, or
799 			 * - ignore the MTU change notification.
800 			 */
801 			icmp6_mtudisc_update((struct ip6ctlparam *)d, valid);
802 
803 			/*
804 			 * regardless of if we called icmp6_mtudisc_update(),
805 			 * we need to call in6_pcbnotify(), to notify path
806 			 * MTU change to the userland (2292bis-02), because
807 			 * some unconnected sockets may share the same
808 			 * destination and want to know the path MTU.
809 			 */
810 		}
811 
812 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6,
813 		    uh.uh_dport, (struct sockaddr *)&sa6_src,
814 		    uh.uh_sport, cmd, cmdarg, notify);
815 	} else {
816 		(void) in6_pcbnotify(&udbtable, (struct sockaddr *)&sa6, 0,
817 		    (struct sockaddr *)&sa6_any, 0, cmd, cmdarg, notify);
818 	}
819 }
820 #endif
821 
822 void *
823 udp_ctlinput(cmd, sa, v)
824 	int cmd;
825 	struct sockaddr *sa;
826 	void *v;
827 {
828 	register struct ip *ip = v;
829 	register struct udphdr *uhp;
830 	extern int inetctlerrmap[];
831 	void (*notify)(struct inpcb *, int) = udp_notify;
832 	int errno;
833 
834 	if (sa == NULL)
835 		return NULL;
836 	if (sa->sa_family != AF_INET ||
837 	    sa->sa_len != sizeof(struct sockaddr_in))
838 		return NULL;
839 
840 	if ((unsigned)cmd >= PRC_NCMDS)
841 		return NULL;
842 	errno = inetctlerrmap[cmd];
843 	if (PRC_IS_REDIRECT(cmd))
844 		notify = in_rtchange, ip = 0;
845 	else if (cmd == PRC_HOSTDEAD)
846 		ip = 0;
847 	else if (errno == 0)
848 		return NULL;
849 	if (ip) {
850 		uhp = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
851 		in_pcbnotify(&udbtable, sa, uhp->uh_dport, ip->ip_src,
852 		    uhp->uh_sport, errno, notify);
853 	} else
854 		in_pcbnotifyall(&udbtable, sa, errno, notify);
855 	return NULL;
856 }
857 
858 int
859 udp_output(struct mbuf *m, ...)
860 {
861 	register struct inpcb *inp;
862 	struct mbuf *addr, *control;
863 	register struct udpiphdr *ui;
864 	register int len = m->m_pkthdr.len;
865 	struct in_addr laddr;
866 	int s = 0, error = 0;
867 	va_list ap;
868 	int pcbflags = 0;
869 
870 	va_start(ap, m);
871 	inp = va_arg(ap, struct inpcb *);
872 	addr = va_arg(ap, struct mbuf *);
873 	control = va_arg(ap, struct mbuf *);
874 	va_end(ap);
875 
876 #ifdef DIAGNOSTIC
877 	if ((inp->inp_flags & INP_IPV6) != 0)
878 		panic("IPv6 inpcb to udp_output");
879 #endif
880 
881 	/*
882 	 * Compute the packet length of the IP header, and
883 	 * punt if the length looks bogus.
884 	 */
885 	if ((len + sizeof(struct udpiphdr)) > IP_MAXPACKET) {
886 		error = EMSGSIZE;
887 		goto release;
888 	}
889 
890 	if (addr) {
891 		/*
892 		 * Save current PCB flags because they may change during
893 		 * temporary connection.
894 		 */
895 		pcbflags = inp->inp_flags;
896 
897 		laddr = inp->inp_laddr;
898 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
899 			error = EISCONN;
900 			goto release;
901 		}
902 		/*
903 		 * Must block input while temporarily connected.
904 		 */
905 		s = splsoftnet();
906 		error = in_pcbconnect(inp, addr);
907 		if (error) {
908 			splx(s);
909 			goto release;
910 		}
911 	} else {
912 		if (inp->inp_faddr.s_addr == INADDR_ANY) {
913 			error = ENOTCONN;
914 			goto release;
915 		}
916 	}
917 	/*
918 	 * Calculate data length and get a mbuf
919 	 * for UDP and IP headers.
920 	 */
921 	M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
922 	if (m == 0) {
923 		error = ENOBUFS;
924 		goto bail;
925 	}
926 
927 	/*
928 	 * Fill in mbuf with extended UDP header
929 	 * and addresses and length put into network format.
930 	 */
931 	ui = mtod(m, struct udpiphdr *);
932 	bzero(ui->ui_x1, sizeof ui->ui_x1);
933 	ui->ui_pr = IPPROTO_UDP;
934 	ui->ui_len = htons((u_int16_t)len + sizeof (struct udphdr));
935 	ui->ui_src = inp->inp_laddr;
936 	ui->ui_dst = inp->inp_faddr;
937 	ui->ui_sport = inp->inp_lport;
938 	ui->ui_dport = inp->inp_fport;
939 	ui->ui_ulen = ui->ui_len;
940 
941 	/*
942 	 * Compute the pseudo-header checksum; defer further checksumming
943 	 * until ip_output() or hardware (if it exists).
944 	 */
945 	if (udpcksum) {
946 		m->m_pkthdr.csum |= M_UDPV4_CSUM_OUT;
947 		ui->ui_sum = in_cksum_phdr(ui->ui_src.s_addr,
948 		    ui->ui_dst.s_addr, htons((u_int16_t)len +
949 		    sizeof (struct udphdr) + IPPROTO_UDP));
950 	} else
951 		ui->ui_sum = 0;
952 	((struct ip *)ui)->ip_len = htons(sizeof (struct udpiphdr) + len);
953 	((struct ip *)ui)->ip_ttl = inp->inp_ip.ip_ttl;
954 	((struct ip *)ui)->ip_tos = inp->inp_ip.ip_tos;
955 
956 	udpstat.udps_opackets++;
957 	error = ip_output(m, inp->inp_options, &inp->inp_route,
958 	    inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST),
959 	    inp->inp_moptions, inp, (void *)NULL);
960 
961 bail:
962 	if (addr) {
963 		in_pcbdisconnect(inp);
964 		inp->inp_flags = pcbflags;
965 		inp->inp_laddr = laddr;
966 		splx(s);
967 	}
968 	if (control)
969 		m_freem(control);
970 	return (error);
971 
972 release:
973 	m_freem(m);
974 	if (control)
975 		m_freem(control);
976 	return (error);
977 }
978 
979 u_int	udp_sendspace = 9216;		/* really max datagram size */
980 u_int	udp_recvspace = 40 * (1024 + sizeof(struct sockaddr_in));
981 					/* 40 1K datagrams */
982 
983 #ifdef INET6
984 /*ARGSUSED*/
985 int
986 udp6_usrreq(so, req, m, addr, control, p)
987 	struct socket *so;
988 	int req;
989 	struct mbuf *m, *addr, *control;
990 	struct proc *p;
991 {
992 
993 	return udp_usrreq(so, req, m, addr, control);
994 }
995 #endif
996 
997 /*ARGSUSED*/
998 int
999 udp_usrreq(so, req, m, addr, control)
1000 	struct socket *so;
1001 	int req;
1002 	struct mbuf *m, *addr, *control;
1003 {
1004 	struct inpcb *inp = sotoinpcb(so);
1005 	int error = 0;
1006 	int s;
1007 
1008 	if (req == PRU_CONTROL) {
1009 #ifdef INET6
1010 		if (inp->inp_flags & INP_IPV6)
1011 			return (in6_control(so, (u_long)m, (caddr_t)addr,
1012 			    (struct ifnet *)control, 0));
1013 		else
1014 #endif /* INET6 */
1015 			return (in_control(so, (u_long)m, (caddr_t)addr,
1016 			    (struct ifnet *)control));
1017 	}
1018 	if (inp == NULL && req != PRU_ATTACH) {
1019 		error = EINVAL;
1020 		goto release;
1021 	}
1022 	/*
1023 	 * Note: need to block udp_input while changing
1024 	 * the udp pcb queue and/or pcb addresses.
1025 	 */
1026 	switch (req) {
1027 
1028 	case PRU_ATTACH:
1029 		if (inp != NULL) {
1030 			error = EINVAL;
1031 			break;
1032 		}
1033 		s = splsoftnet();
1034 		error = in_pcballoc(so, &udbtable);
1035 		splx(s);
1036 		if (error)
1037 			break;
1038 		error = soreserve(so, udp_sendspace, udp_recvspace);
1039 		if (error)
1040 			break;
1041 #ifdef INET6
1042 		if (((struct inpcb *)so->so_pcb)->inp_flags & INP_IPV6)
1043 			((struct inpcb *) so->so_pcb)->inp_ipv6.ip6_hlim =
1044 			    ip6_defhlim;
1045 		else
1046 #endif /* INET6 */
1047 			((struct inpcb *) so->so_pcb)->inp_ip.ip_ttl = ip_defttl;
1048 		break;
1049 
1050 	case PRU_DETACH:
1051 		udp_detach(inp);
1052 		break;
1053 
1054 	case PRU_BIND:
1055 		s = splsoftnet();
1056 #ifdef INET6
1057 		if (inp->inp_flags & INP_IPV6)
1058 			error = in6_pcbbind(inp, addr);
1059 		else
1060 #endif
1061 			error = in_pcbbind(inp, addr);
1062 		splx(s);
1063 		break;
1064 
1065 	case PRU_LISTEN:
1066 		error = EOPNOTSUPP;
1067 		break;
1068 
1069 	case PRU_CONNECT:
1070 #ifdef INET6
1071 		if (inp->inp_flags & INP_IPV6) {
1072 			if (!IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1073 				error = EISCONN;
1074 				break;
1075 			}
1076 			s = splsoftnet();
1077 			error = in6_pcbconnect(inp, addr);
1078 			splx(s);
1079 		} else
1080 #endif /* INET6 */
1081 		{
1082 			if (inp->inp_faddr.s_addr != INADDR_ANY) {
1083 				error = EISCONN;
1084 				break;
1085 			}
1086 			s = splsoftnet();
1087 			error = in_pcbconnect(inp, addr);
1088 			splx(s);
1089 		}
1090 
1091 		if (error == 0)
1092 			soisconnected(so);
1093 		break;
1094 
1095 	case PRU_CONNECT2:
1096 		error = EOPNOTSUPP;
1097 		break;
1098 
1099 	case PRU_ACCEPT:
1100 		error = EOPNOTSUPP;
1101 		break;
1102 
1103 	case PRU_DISCONNECT:
1104 #ifdef INET6
1105 		if (inp->inp_flags & INP_IPV6) {
1106 			if (IN6_IS_ADDR_UNSPECIFIED(&inp->inp_faddr6)) {
1107 				error = ENOTCONN;
1108 				break;
1109 			}
1110 		} else
1111 #endif /* INET6 */
1112 		{
1113 			if (inp->inp_faddr.s_addr == INADDR_ANY) {
1114 				error = ENOTCONN;
1115 				break;
1116 			}
1117 		}
1118 
1119 		s = splsoftnet();
1120 		in_pcbdisconnect(inp);
1121 #ifdef INET6
1122 		if (inp->inp_flags & INP_IPV6)
1123 			inp->inp_laddr6 = in6addr_any;
1124 		else
1125 #endif /* INET6 */
1126 			inp->inp_laddr.s_addr = INADDR_ANY;
1127 
1128 		splx(s);
1129 		so->so_state &= ~SS_ISCONNECTED;		/* XXX */
1130 		break;
1131 
1132 	case PRU_SHUTDOWN:
1133 		socantsendmore(so);
1134 		break;
1135 
1136 	case PRU_SEND:
1137 #ifdef INET6
1138 		if (inp->inp_flags & INP_IPV6)
1139 			return (udp6_output(inp, m, addr, control));
1140 		else
1141 			return (udp_output(m, inp, addr, control));
1142 #else
1143 		return (udp_output(m, inp, addr, control));
1144 #endif
1145 
1146 	case PRU_ABORT:
1147 		soisdisconnected(so);
1148 		udp_detach(inp);
1149 		break;
1150 
1151 	case PRU_SOCKADDR:
1152 #ifdef INET6
1153 		if (inp->inp_flags & INP_IPV6)
1154 			in6_setsockaddr(inp, addr);
1155 		else
1156 #endif /* INET6 */
1157 			in_setsockaddr(inp, addr);
1158 		break;
1159 
1160 	case PRU_PEERADDR:
1161 #ifdef INET6
1162 		if (inp->inp_flags & INP_IPV6)
1163 			in6_setpeeraddr(inp, addr);
1164 		else
1165 #endif /* INET6 */
1166 			in_setpeeraddr(inp, addr);
1167 		break;
1168 
1169 	case PRU_SENSE:
1170 		/*
1171 		 * stat: don't bother with a blocksize.
1172 		 */
1173 		/*
1174 		 * Perhaps Path MTU might be returned for a connected
1175 		 * UDP socket in this case.
1176 		 */
1177 		return (0);
1178 
1179 	case PRU_SENDOOB:
1180 	case PRU_FASTTIMO:
1181 	case PRU_SLOWTIMO:
1182 	case PRU_PROTORCV:
1183 	case PRU_PROTOSEND:
1184 		error =  EOPNOTSUPP;
1185 		break;
1186 
1187 	case PRU_RCVD:
1188 	case PRU_RCVOOB:
1189 		return (EOPNOTSUPP);	/* do not free mbuf's */
1190 
1191 	default:
1192 		panic("udp_usrreq");
1193 	}
1194 
1195 release:
1196 	if (control) {
1197 		m_freem(control);
1198 	}
1199 	if (m)
1200 		m_freem(m);
1201 	return (error);
1202 }
1203 
1204 static void
1205 udp_detach(inp)
1206 	struct inpcb *inp;
1207 {
1208 	int s = splsoftnet();
1209 
1210 	in_pcbdetach(inp);
1211 	splx(s);
1212 }
1213 
1214 /*
1215  * Sysctl for udp variables.
1216  */
1217 int
1218 udp_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
1219 	int *name;
1220 	u_int namelen;
1221 	void *oldp;
1222 	size_t *oldlenp;
1223 	void *newp;
1224 	size_t newlen;
1225 {
1226 	/* All sysctl names at this level are terminal. */
1227 	if (namelen != 1)
1228 		return (ENOTDIR);
1229 
1230 	switch (name[0]) {
1231 	case UDPCTL_CHECKSUM:
1232 		return (sysctl_int(oldp, oldlenp, newp, newlen, &udpcksum));
1233 	case UDPCTL_BADDYNAMIC:
1234 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1235 		    baddynamicports.udp, sizeof(baddynamicports.udp)));
1236 	case UDPCTL_RECVSPACE:
1237 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_recvspace));
1238 	case UDPCTL_SENDSPACE:
1239 		return (sysctl_int(oldp, oldlenp, newp, newlen,&udp_sendspace));
1240 	default:
1241 		return (ENOPROTOOPT);
1242 	}
1243 	/* NOTREACHED */
1244 }
1245